Dependencies: Permit installation on Python 3.13

[amotl]

Please do not merge.

About

The patch satisfies the installation of PyCaret on Python 3.13 without needing a compiler. It does not validate relevant functionality.

uv pip install 'pycaret @ git+https://github.com/sktime/pycaret@python313'

Thoughts

The patch does not take care about actually resolving potential havocs on individual version updates yet. It is meant to use the branch to quickly validate it on downstream CI jobs, and to convey an overview about what needs to be addressed.

In this spirit, it certainly needs more love and validation, optimally on behalf of individual patches that address individual package updates.

NB: This patch builds upon @fkiraly's other one which relaxes the dependency on pmdarima. Thank you very much.

• [MNT] isolate pmdarima dependency pycaret/pycaret#4129

[amotl]

Do you know why CI is not starting on this PR, @fkiraly?

[fkiraly]

it is starting (and failing) - why do you believe it is not starting?

[fkiraly]

The failure is likely since you changed to many things at once, e.g., upper bounds for various packages.

I think the problem comes from numpy 2.

I suggest we raise the bounds one-by-one, or it will be really hard to diagnose and fix.

[amotl]

After pushing once more, CI started running again, so I guess it was just a hiccup on GHA, possibly just running out of runners in peak hours.

[amotl]

The failure is likely since you changed to many things at once, e.g., upper bounds for various packages.

I don't know how this would contribute to CI not starting at all?

I think the problem comes from numpy 2.
I suggest we raise the bounds one-by-one, or it will be really hard to diagnose and fix.

I absolutely agree. This patch was not meant to address individual problems, just make the installation of PyCaret on Python 3.13 succeed, in order use the branch to quickly validate it on downstream CI jobs. I've adjusted the original post to reflect its purpose, and that it should not be merged.

[fkiraly]

I don't know how this would contribute to CI not starting at all?

Probably some dependency conflicts.

[amotl]

I think it was just a hiccup on GHA, at the time I was pushing to the branch.